The 245 bacterial, 26 archaeal and 49 eukaryote genomes, and the identified known and putative globins, are listed in Supplemental Data Tables 1–6 [see Additional File 1]. An alignment of the sequences is provided in Supplemental Data Fig. 1 [see Additional File 2]. All the globins identified in our survey can be assigned to two general classes: single chain globins and globin domains within chimeric proteins. The former group comprises the vertebrate α- and β-globins, Mbs, Ngbs and Cygbs, the metazoan intra- and extracellular Hbs (including multi-domain and/or multisubunit Hbs), the plant SHbs and NsHbs, the ~195aa Pgbs and the 2/2 Hbs. The two groups of chimeric proteins, which generally have an N-terminal globin domain, are the ~400aa FHbs with an NAD and FAD-binding C-terminal domain, and the GCSs with highly variable (~50 to >700a) C-terminal domains. Thus, all globins belong to one of three globin lineages: the 3/3 FHbs and SDgbs (eukaryote and bacterial), the 3/3 GCSs and Pgbs and the 2/2 SDgbs 49. Fig. 1 provides the number of genomes that have globins and illustrates a key result: only one of the three globin lineages, the 2/2 Hbs, are represented in all three kingdoms of life. Fig. 2 shows the approximate phylogenetic relationships between the main groups of the three kingdoms of life based on Baldauf et al. 50 and depicts the phylogenomic profile of globins.

It should be emphasized, that the presence of putative globins inferred using SUPERFAMILY and blastp and psiblast searches of the GenBank database, does not require having a completed genome. On the other hand, the absence of a globin is certain only when a complete genome is searched and no sequence is found.

Table 2 in the Supporting Data [see Additional File 1] lists the archaeal genomes, their sizes and the presence of globins. SUPERFAMILY identified putative globins in four Archaea: Halobacterium, Methanobacterium, Methanosarcina and Sulfolobus. In addition, another globin was identified in the SWISSPROT/TrEMBL database (tr|Q96ZL6). Examination of these sequences showed that they are not likely to be globins due to lack of size (< 100aa) and/or the absence of a suitable F8His. However, Hou et al. [44.45,48] have recently discovered GCSs and Pgbs in a couple of archaea, and we found a 2/2 Hb1 in Haloarcula marismortui 49. This survey has uncovered a 2/2 Hb1 in still another euryarcheote halophile Haloferax volcanii. Thus, only five of the 20 Euryarcheote genomes have globins: Haloarcula marismortui (2/2 Hb1, GCS), Halobacterium salinarum (GCS), Haloferax volcanii (2/2 Hb1, GCS), Methanosarcina acetivorans and M. barkeri (Pgbs). Of the 5 Crenarcheote genomes, only Aeropyrum pernix, has a globin (Pgb) and the only Nanoarcheote genome has no globins. The paucity of globins in Archaea is one of the surprising results of our survey. Except for Aeropyrum pernix, the archaeal genomes comprising globins are, coincidentally, among the largest known archaeal genomes. It is interesting to note that none of the GCSs or Pgbs have been identified by the hidden Markov chain model used by SUPERFAM, although they align readily with the globin fold, based on the recent crystal structure of Bacillus subtilis GCS 51. Most of the archaeal genomes sequenced to date have been those of extremophilic organisms with small genomes. It is now known that there are many more non-thermophilic terrestrial and marine archaea than hitherto suspected 52. It remains to be seen whether additional archaeal genomes will reverse the situation.

Table 3 in Supporting Data [see Additional File 1] lists the completed and unfinished bacterial genomes used in this study, their sizes, the type and number of globins present, 2/2 Hbs (classes 1, 2 and 3), FHbs (SDgbs) and GCSs (Pgbs), and the oxygen requirement and habitat of the organism. A diagrammatic representation of the phylogenetic relationships between the main bacterial groups based on Baldauf et al. 50, shown in Fig. 3, summarizes the globin distribution among the bacteria. The one representative of the Acidobacteria has globins, and only 4 of the 26 genomes representing the Actinomycetes have none: Bifidobacterium, Propionibacterium, Symbiobacterium and Trophyrema. The 3 Aquificales/Thermotogales have globins except Thermotoga, and the 5 Bacteroidetes/Chlorobi (except Cytophaga) lack globins, as well as the 6 Chlamydiae/Verrumicrobium genomes, except Parachlamydia and Verrumicrobium. Of the 3 Chloroflexi genomes, Dehalococcoides, has no globins. The Cyanobacteria are represented by 12 genomes, of which 5 lack globins: Anabaena, Crocosphaera, Prochlorococcus, Synechocccus elongatus and Trichodesmium. Both Deinococcus and Thermus have globins. Among the 49 Firmicutes, apart from the Lactobacillales and the Mollicutes, the 17 Bacilli have globins (except Listeria), as well as 4 of the 7 Clostridia. Although Fusobacterium lacks globins, the lone Nitrospirae and the two Planctomycetes have globins. The majority of the 35 genomes of Alphaproteobacteria have globins: all the Caulobacterales, Rhizobiales (except Bartonella), Rhodobacterales, Rhodospirilalles and Sphingomonadales, but only one of the 10 Rickettsiales (Pelagibacter ubique). All 24 genomes representing the Betaproteobacteria have globins, except 2 of the 4 Neisseriaceae. The 50 Gammaproteobacteria genomes represent 10 groups, of which only 3 of the 10 Enterobacteriales (Buchnera, Blochmannia and Wigglesworthia), one of two Legionellales, the 5 Pasteurellales, the two Psychrobacter of the 10 Pseudomonadales, and the one Thiotrichales (Francisella) are devoid of globins. The 10 genomes of Deltaproteobacteria have globins, except for Desulfovibrio. Two of the 8 genomes representing the Epsilonproteobacteria have no globins: Helicobacter pylorii and Wolinella; however, Helicobacter hepaticus has a globin. The lone unassigned proteobacterium Magnetococcus has a GCS. Of the 128 genomes representing the Proteobacteria, 99 have globins (~75%).

Alignment of bacterial 2/2 Hb sequences indicates that they can be divided into three separate classes 1653. The length of the 2/2 Hbs varies from 118aa for the cyanobacterium Nostoc (ZP_00112318) to 260aa in Streptomyces avermitilis (NP_824492), with the majority (>60) being <140aa. Examination of coexistence of the three classes of 2/2 Hbs (Table 1) shows an interesting trend: although only the Alphaproteobacterium Hyphomonas neptunium has both 1 and 3 and two have all three classes, 14 genomes have classes 1 and 2 and 18 have classes 2 and 3. These results suggest that the class 1 and 3 2/2 Hbs are derived from the class 2 2/2 Hbs, as pointed out very recently by Vuletich and Lecomte 53.

The finding of an FHb in E. coli 7 led to the discovery of some 30 bacterial FHbs 16424354; currently, the number is over 70 (Supplementary Data Table 3). R. Poole and his group 43 and Frey and Kallio 42 were the first to point out the similarity between the FHb globin domains, the first bacterial Hb to be sequenced, that of Vitreoscilla stercoraria 55, and several other SD bacterial globins. We now have a total of 25 SDgbs in 22 genomes. The crystal structures of the P. aeruginosa (PDB: 1tu9) 56 and Vitreoscilla stercoraria (PDB: 2vhb) 57, show them to be very similar to the globin domains of the FHbs from E. coli (PDB: 1gvh) 58 and Ralstonia (Alcaligenes) eutropha (PDB: 1cqx) (59]. It is appropriate to note here that Ralstonia has undergone two name alterations recently, to Wautersia eutropha and now to Cupriavidus necator 60. Although 22 genomes have SDgbs, three of them have two different globins: Bradyrhizobium, Rhodopseudomonas and Novosphingobium. Table 2 shows that FHbs and SDgbs coexist in only 6 genomes: Chromobacterium, Photobacterium, Pseudomonas aeruginosa, Rhodopirellula, Thermobifida and Vibrio parahaemolyticus. Table 3 shows the statistics of occurrence of 2/2 Hbs with the FHbs/SDgbs; the latter tend to occur with class 2 and with both class 2 and 3 2/2 Hbs.

M. Alam and his group have identified 27 GCSs 4647 and 3 Pgbs from bacteria – Chloroflexus aurantiacus ZP_00359040 (227aa), Thermobifida fusca ZP_00293478 (197aa) and Thermosynechococcus elongatus NP_682779 (194aa) 48. Blastp searches found an additional two dozen GCSs, and Pgbs in Thermus thermophilus YP_005074 (203aa) and the actinobacterium Rubrobacter xylanophilus ZP_00200180 (196aa) 4961. The recent crystal structure of B. subtilis GCS shows it to have a 3/3 fold 62.

The genomes with the largest number of globins, 5 to 7, are all from the Alpha- and Betaproteobacteria; they do include however, redundant sequences. The following have all three lineages of globins:Burkholderia fungorum (9.67 Mbp), Chromobacterium violaceum (4.75 Mbp), Novosphingobium aromaticivorans (4.21 Mbp) and Silicibacter sp.TM1040 (4.14 Mbp), each with 5 globins, the three Bordetella species (4.09–5.34 Mbp), Exiguobacterum sp.255–15 (2.89 Mbp), Shewanella baltica (5.01 Mbp), Sinorhizobium meliloti and Thermobifida fusca (3.64 Mbp), each with 4 globins, and Azotobacter vinelandii (5.42 Mbp), all the Bacillus species (except B. licheniformis and B. stearothermophilus) (4.20–5.50 Mbp) and Geobacillus kaustophilus (3.59 Mbp), each with 3 globins. At the other end of the scale, the smallest genome (1.31 Mbp) to have a globin, a 2/2 Hb1, is that of the marine Alphaproteobacterium Pelagibacter ubiques (Rickettsiales), followed by Aquifex aeolicus (1.59 Mbp), and the Epsilonbacterium Campylobacter jejuni (1.64 Mbp), with one (SDgb) and two (2/2Hb and SDgb) globins, respectively.

The bacterial divisions represented by 10 or more genomes, rank in the following order of having globins: Betaproteobacteria – 92% (22/24) > Actinobacteria – 85% (22/26) > Gammaproteobacteria – 76% (38/50) > Alphaproteobacteria – 69% (24/35) > Cyanobacteria – 58% (7/12) > Firmicutes – 49% (19/39). Overall, 161 of the 245 (~65%) bacterial genomes have globins and the number of globins varies from 1 to 7.

Fig. 4 depicts the size distribution of bacterial genomes lacking globins (upper panel) and comprising globins (lower panel); the stippled columns represent the genomes associated with a host (given in Supplemental Data Table 3 [see Additional File 1]). Although the two distributions overlap, it is evident that presence of globins is correlated with genome size and that the majority of genomes lacking globins are host associated. Table 4 shows the mean genome size for completed genomes lacking globin and with one or more globins. The mean size of the genomes shows a positive correlation with the number of globins present. About 80% (65/82) of genomes <2.5 Mbp lack globins (mean 2.1 Mbp). Very small genomes are characteristic of bacteria with lifestyles characterized by continuous association with a host 63. The majority of the bacteria whose genomes have been sequenced, are human, animal and plant infectious agents. Thus, it is likely that the estimate of approximately two thirds of all bacteria having globins obtained in the present study, is on the low side.

The globins identified in eukaryote genomes are listed in Supplemental Data Tables 4–6 [see Additional File 1]. The globins listed for the vertebrate genomes, Danio (Brachydanio) rerio, Fugu (Takifugu) rubripes, Gallus gallus, Homo sapiens, Mus musculus, Rattus norvegicus and Xenopus tropicalis, include the familiar α- and β-globins and Mb, as well as the recently discovered Ngbs 3334 and Cygbs 353637. A recent addition to human α-globins is μ-globin (gi|51510893|NP_001003938, 141aa) which appears to be very similar to the avian α-D globin 64. In addition, an eye-specific globin, GbE, was found in Gallus gallus 39 and a new globin, GbX, was discovered in the fish, bird and amphibian genomes 40. Curiously, no Mb-like sequence was found in either Xenopus tropicalis or X. laevis. The zebrafish Danio rerio has 6 embryonic globins 65, three α and three β- only 5 are listed in the GenBank. The pufferfish Fugu rubripes, appears to have four α-globin chains 66; only three of them are listed in the GenBank.

The four globins found in the genome of the urochordate Ciona intestinalis (sea squirt), have been identified earlier 32. This important finding corrects the erroneous claim of total absence of globins in the report of the genome 67.

At least one putative globin (gi|72169631|XP_795670, 175aa) was found in the recently completed genome assembly of the echinoderm Strongylocentrotus purpuratus (sea urchin). Used as query in blastp searches it hits the GlbXs and Cygbs of fish, and other vertebrate Cygbs, before recognizing the coelomic globin (gi|729576;P80018) of a fellow echinoderm, the holothurian Caudina arenicola.

There appear to be 4 putative globins in Anopheles gambiae: gi|55246163| EAA03862.3 (150aa), gi|31201271|XP_309583 (192aa), gi|31198253|XP_308074 (215aa) and gi|57914327|XP_555006 (182aa). Their alignment shows that the first three are closely related, sharing essentially identical globin domains: gi|55246163 has an 18aa deletion in the BC helix and the CD corner, while gi|31198253 lacks part of the G and all of the H helices. Thus, it is likely that gi|31201271 and gi|57914327 are the only stable globins. The C-terminal 115aa globin domain of gi|31198253 is reminiscent of the Mb gene sequence from the antarctic icefish Champsocephalus esox (gi|24266940|AAN52371, 103aa) that does not appear to be expressed 68. It remains to be determined whether the N-terminal domain of gi|31198253 stabilizes it enough to be expressed.

At least 33 putative globins and globin domains were identified in Caenorhabditis elegans, together with their orthologs in C. briggsae 30: they are listed in Supplemental Data Table 5 [see Additional File 1]; their alignment is provided in Supplemental Data Fig. 2 [see Additional File 2]. Several C. elegans/C. briggsae orthologs have unusually long interhelical inserts between the G and H helices: 40aa in CE01012/CBP03870) (216aa), 21aa in CE01528 (CBP00622) (266aa), 22aa in CE34964/CBP23619 (196aa) and 21aa in CE34658/CBP07299 (230aa). Although three pairs, CE17437/CBP02580, CE04582/CBP03989 and CE35828/CBP14482, have N-terminal globin domains, several have C-terminal domains: CE03523/CBP15914, CE04843/CBP01576, CE05316/CBP02907, CE12774/CBP21478), CE36044/CBP02078), CE29586/CBP02293, CE30683/CBP02390 and CE31132/CBP15597. The N-terminal portion of the latter is identified as a G protein-coupled receptor-like domain with 7 putative transmembrane helices (identified using TMHMM Server V.2.0). The nonglobin domains in the remaining proteins could not be identified via blastp searches.

Although globin genes were thought to be absent in the genome of Drosophila melanogaster 69, the presence of at least one (CG9734) has been demonstrated unequivocally by Burmester and Hankeln 3170. There appear to be two more putative globins, CG15180 (209aa) and CG14675 (195aa); both provide acceptable alignments. Three putative globins are also found in D. pseudoobscura, EAL28410 (152aa), EAL28093 (641aa, 40–150) and EAL 28094 (130aa) 71, and appear to be orthologs of D. melanogaster CG9734, CG15180 and CG14675, respectively. Although the D. pseudoobscura EAL28093 and EAL28094 are scored by FUGUE as certainly globins (Z ~10), the former lacks an appropriate F8 His residue and the latter appears to be missing the A helix. Psiblast searches show that the ortholog pairs CG15180/EAL28093 and CG14675/EAL28094) recognize each other but not the CG9734/EAL28410 pair, and also recognize the Anopheles globins (ENSANG19788, 22287 and 26474), together with distant recognition of echinoderm globins and vertebrate β-globins. The CG9734/EAL28410 pair hits the other known insect globins (Gasterophilus, Chironomus species, Kiefferulus, Tokunagayusurika), the vertebrate Cygbs and the arthropod multidomain Hbs.

Three of the four Arabidopsis thaliana sequences, GLBs 1–3 have been identified earlier 1972: GLB1 and GLB2 are NsHbs and GLB3 is a 2/2Hb. A 693aa protein (gi|7486404|T04457) was found to have an N-terminal 2/2Hb domain, with the first 138aa (up to position H13), identical to GLB3. Although the ~500aa C-terminal portion does not correspond to any known protein domain(s), a blastp search using it as query found an identical sequence in the N-terminal portion of the 2154aa protein TEBICHI (gi|62241195|BAD93700). Both proteins share the same region on Arabidopsis chromosome 4 and TEBICHI has a helicase domain at 490–890aa and a polymerase domain at 1700–2150aa; furthermore, this N-terminal portion is missing in animal homologues of TEB protein, MUS308/POLQ (Dr. Soichi Inagaki, personal communication).

The genome of Oryza sativa contains the four NsHbs identified by Arredondo-Peter et al. 7374, a 172aa 2/2Hb (gi|50725383|BAD32857) and a 145aa globin (gi|50932383|XP_475719). A blastp search identified the latter as another NsHbs.

Several putative globins occur in the genome of the green alga Chlamydomonas reinhardtii: 160981 (C_240146) (136aa), 160982 (C_240147) (147aa), 157690 (C_1780015) (231aa), and two larger sequences, 168934 (C_60169) (476aa) with one N-terminal globin-like domain, and 153190 (C_100138) (837aa) with two consecutive N-terminal globin domains. Two globins were found in the genome of the diatom Phaeodactylum tricornutum: PTMM3909 and 05212. Two globins occur in the genome of another diatom Thalassiosira pseudonana, Scaffold_137 (158aa) and Scaffold_18 (160aa), and a single 185aa globin, CMR319C, in the genome of the red alga Cyanidioschyzon merolae. Blastp searches identified all the C. reinhardtii globins, the Phaeodactylum PTMM 05212 and the Thalassiosira Scaffold_18 as 2/2 Hbs, and the Cyanidioschyzon CMR319C,Phaeodactylum PTMM 3909 and Thalassiosira Scaffold_137 as SDgbs, closely related to the bacterial and fungal FHbs and the bacterial SDgbs.

The globins present in the fungi are listed in Supplemental Data Table 6 [see Additional File 1]. All the recently completed fungal genomes, which belong overwhelmingly to the Ascomycota (17 of 19), have FHbs; 14 have two to four FHbs, probably due to genome duplication 75. In addition to Saccharomyces cerevisiae NP_014165 (426aa, 154–302) which was known to have a central globin domain 76, we found an additional 12 FHbs to have globin domains within the central portion of their sequences. All belong to the Saccharomycotina: Candida albicans EAK92722 (563aa, 298–463), Candida glabrata XP_448033 (432aa, 124–267), Eremothecium (Ashbya) gossypii NP_982746 (436aa, 184–339), Kluyveromyces lactis XP_453939 (430aa, 181–323), Kluyveromyces waltii Kwal_22190 (421aa, 171–314) Kwal_4395 (460aa, 202–352) and Kwal_24852 (543, 173–315), Saccharomyces bayanus ORFP:20532(424aa, 171–303), Saccharomyces mikatae ORFP:18051 (410aa, 135–290), Saccharomyces paradoxus ORFP:18484 (426aa, 155–306) and Yarrowia lipolytica XP_502881 (463aa, 186–325) and XP_499869 (471aa, 194–333). Although the globin domains align well with the other FHbs, the N- and C-terminal portions do not recognize the C-terminal moieties of the other fungal and bacterial FHbs or any other known protein, as found earlier for the S. cerevisiae FHb 76. Surprisingly, the percent identities between the CDFHbs and the fungal and bacterial FHb globin domains is about 40% and 30%, respectively.

Among the lower eukaryotes, the diplomonad Giardia lamblia has one FHb 77 and the mycetozoan Dictyostelium discoideum has two 78. The Giardia sequence has a 21aa insert between helices E and F.

No globin genes appear to exist in the genomes of the Apicomplexans Entamoeba histolytica and Plasmodium falciparum, as well as the Microsporidian Encephalitozoon cuniculi and the Euglenazoan Trypanosoma brucei. Although a Plasmodium bergei globin-like protein of 228aa (Q86QI8) is listed in the GenBank, it is probably an artefact, since when used as a query in blastp searches, it hits only vertebrate β- and α-globins.

Compared to Archaea and Bacteria, the fraction of eukaryote genomes with globins is much higher, over 90%. Although all vertebrates have Hb and Mb and probably Ngb and Cygb as well 38, the Antarctic icefish belonging to the family Channichthyidae do not express Hb 79. Furthermore, at least 6 of the 16 icefish species also do not express Mb 80. The lack of Hb is due apparently to deletion of the β-globin gene 81, which occurred during the last 10 to 16 Myr, the approximate date of divergence of the Notothenioid lineage including the Channichthyidae 82. In contrst, the lack of Mb appears to be due to errors in Mb gene transcription 68. It is not known whether the icefish have Ngbs and/or Cygbs. In the remaining two chordate phyla, the urochordate Ciona has 4 globins (see above), which when used as queries in psiblast searches, recognize the vertebrate globins and the eukaryote and bacterial FHbs and SDgbs. The report of a Hb in the notochord of the cephalochordate amphioxus (Branchiostoma californiense) 83, suggests that globins are present in all chordates. Among the remaining deuterostome phyla, intracellular Hbs have been reported in two of the 5 classes of echinoderms 8485, and we find a putative globin in the recent assembly of the genome from the sea urchin Strongylocentrotus purpuratus, which belongs to another class. No globins have been reported so far in hemichordates.

Among the remaining metazoans, direct visual observation of Hbs is highly episodic among the metazoans, and varies from total absence of Hb in porifera and cnidaria, to frequent presence in some nematode and platyhelminth groups 3. In the largest metazoan group, the insects, of which about 900,000 species have been described, and which outnumber the combined total of all other animal species, with estimates ranging from 2 × 10⁶ to 100 × 10⁶ 86, globins are so far known to be present only in Dipterans, such as Gasterophilus, Drosophila and Chironomus. It is evident that much work needs to be done before we have any clear idea of globin distribution in metazoans other than deuterostomes.

Among the lower eukaryotes, our census suggests that globins could be ubiquitous in fungi (FHbs), mycetozoa (FHbs), diplomonads (FHbs), ciliates (2/2 Hbs), stramenopiles (SDgb, 2/2Hb), rhodophytes (SDgb), chlorophytes (2/2Hb) and plants (SHbs, NsHbs, 2/2 Hbs). In contrast, the genomes of pathogenic microsporidians, entamoebae, apicomplexans anad trypanosomes are devoid of globins. In view of a very large potential diversity of small, bacterial-sized eukaryotes revealed by recent, culture-independent surveys 8788, much remains to be done.

Since our census indicates that GCSs do not occur in eukaryotes, it is necessary to consider the heme-regulated eukaryotic initiation factor 2α kinase (~630aa), which was found recently to have two heme-binding domains, of which the N-terminal domain was considered to be globin-like 8990 Although it was demonstrated that His78 and His123 were the two residues involved in heme binding 91, we find that the N-terminal 138aa of the rabbit protein (gi|462439|P33279|E2AK1 RABIT) is not recognized as a globin in CD or FUGUE searches. A blastp search of the GenBank database found at least 6 mammalian counterparts; clearly, a structure is required to determine whether these sequences are globins.

The broad distribution of the three lineages of globins, the 3/3 FHbs/SDgbs, the 3/3 GCS/Pgbs and the 2/2 Hbs, is illustrated in the Bayesian phylogenetic tree shown in Fig. 5, based on a global manual alignment of 175 sequences representing all the major groups of globins (Supplemental Data Fig. 3, [see Additional File 2]). The 2/2 Hbs and the GCSs/Pgbs cluster separately (upper left hand side), with the three classes of 2/2 Hbs clearly delineated. Within the 2/2Hb clusters, the archaeal (Haloarcula), ciliate (Paramecium) and chlorophyte (Chlamydomonas) globins occur with the class 1 and the diatom (Thalassiosira) and plant (Oryza, Arabidopsis) Hbs group with the class 2. The remaining stem encompasses the bacterial and eukaryotic FHbs and related SDgbs (marked with an asterisk), including the Cyanidioschyzon and Thalassiosira SDgbs, the separate clusters of plant NsHbs and SHbs (lower right hand side) and all the known metazoan globins, including the vertebrate globins (lower left hand side). Noteworthy is the clustering of bacterial SDgbs with the globin domains of bacterial and eukaryote (fungal, diplomonad and mycetozoan) FHbs, suggestive of past lateral gene transfer events 49. Likewise, the clustering of the archaeal 2/2 Hbs with the class 1 bacterial 2/2 Hbs, could be indicative of a past lateral gene transfer. Very recently, a 2/2Hb and an FHb were found in the multicellular red alga Chondrus crispus (X. Bailly et al., unpublished results): these two sequences cluster with the Thalassiosira 2/2Hb and Aquifex SDgb, respectively. An interesting result is the close grouping of Amphitrite ornata dehaloperoxidase 92 with the intracellular hemoglobin of the hydrothermal vent annelid Alvinella pompejana 93, suggesting that the latter may be a dehaloperoxidase also. The Gallus and Xenopus HbX group with the vertebrate Ngbs as found earlier 40. Of particular interest is the large C. elegans globin family: all except F21A3.6 and ZK637.13, cluster together, and appear to be closest to the vertebrate Ngbs. ZK637.13 groups with globins from other nematode species, F21A3.6 clusters with crustacean (Daphnia and Artemia) and platyhelminth (Paramphistomum and Clonorchis) sequences. It should be noted that the locations of these three globins is dependent on the number of globin sequences used to construct the phylogenetic tree.

We have discussed elsewhere 49 the possible evolutionary scenarios for the emergence of the proposed three lineages of globins in two structural classes: (1) the 3-over-3 SD globins and FHbs, (2) the 3-over-3 GCS/Pgbs, and (3), the 2-over-2 SDgbs. Here, we would like to reiterate our proposal that all metazoan and plant globins originated from a SDgb related to the present day bacterial and algal SDgbs and the N-terminal globin domains of fungal and bacterial FHbs. This proposal is based on two results, one of which, is the clustering of all metazoan and plant sequences with the bacterial and eukaryote FHbs and SDgbs and separate from the branches encompassing 2/2 Hbs and GCSs/Pgbs, in the Bayesian phylogenetic tree obtained earlier 49 and in the one shown in Fig. 5. Another key result, is that bacterial SDgbs and eukaryote FHb globin domains used as queries in iterated psiblast searches, recognize (i.e. have E values substantially lower than threshold) vertebrate globins, particularly Ngbs, and other metazoan globin groups, ahead of the 2/2 Hbs and GCSs/Pgbs 49. Conversely, vertebrate Ngbs as a group, recognize the bacterial SDgbs ahead of other vertebrate globins. Furthermore, the tree shown in Fig. 5 is in very good agreement with recent models of vertebrate globin evolution 366294, wherein duplication of the ancestral globin gene resulted in neuroglobin and cellular globin genes, with subsequent duplication of the latter into cellular and Hb gene loci, followed by additional duplications of the cellular globin locus into Mb and Cygb and of the Hb gene into the α- and β-globin genes.

The first attempt at defining the molecular phylogeny of globins 95 was a tree based on 245 sequences, all from eukaryotes. A subsequent study by Moens et al. 96 based on 700 sequences, including bacterial ones, proposed that all globins «evolved from a family of ancestral, ca. 17 kDa hemeproteins, which displayed the globin fold and functioned as redox proteins». The present survey of extant globins in the three kingdoms of life provides some interesting additional details: we now know that there are three lineages of globins, two with a 3/3 α-helical fold and one with the 2/2 fold, and that only the latter occurs at present in all three kingdoms of life. Furthermore, it appears likely that formation of chimeric proteins containing globin domains occurred, as separate events, prior to the emergence of the three kingdoms of life. There are no obvious explanations for the absence of FHbs in Archaea and of GCSs in eukaryotes. Although it is not possible at present to decide which of the two folds originated from the other, the occurrence of 2/2 Hbs in all three kingdoms of life would suggest that it is the ancestral fold. A recent computational study of plant Hb folding showed that one of the folding modules of rice 3/3 NsHb overlaps the 2/2 fold of Mycobacterium tuberculosis HbO (class 2) suggesting that it is an ancient structural feature of globins 97.

The pesence of multiple globins in all three kingdoms of life raises the question of their function, paticularly among the unicellular organisms. Although a complete discussion is not possible here, some general observations can be made, limited to organisms other than Animalia. The bacterial FHbs appear to have a role primarily in the detoxification of NO, via two NADH dependent activities, an NO dioxygenase activity under aerobic conditions, or an NO reductase activity under anaerobic conditions 42439899100. Furthermore, the recent demonstration that the E. coli FHb binds lipids and is an efficient alkylhydroperoxide reductase, suggests that it may be involved in the repair of lipid membranes damaged by oxidative/nitrosative stress 101102103. The function(s) of bacterial SDgbs appear to be somewhat similar to the FHbs. Although the function of Vitreoscilla Hb is to increase the effective intracellular O₂concentration under microaerobic conditions 42104, that of Campylobacter SDgb was shown recently to function only in NO scavenging and detoxification and not provide resistance against superoxide or peroxides 105. In eukaryotes, such as yeasts and Dictyostelium, FHbs also provide protection against NO 4310678, as well as enhancement of respiration, directly by functioning as an O₂ buffer and indirectly, by reducing the NO concentration in the mitochondria which otherwise inhibits respiration 107108109110. In pathogenic microorganisms, FHb provides protection from human macrophage NO-mediated killing and promotes the virulence of bacteria, e.g. Salmonella 111, and of yeasts, Candida 112 and Cryptococcus neoformans, a worldwide pathogen causing pulmonary infection in animals and humans 113114. It is interesting to note, as suggested by one of the reviewers, that since most of the Archaea are not pathogenic, they may have dispensed with the FHbs and their protective role in bacteria.

In plants, the SHbs have been shown to be required for establishing a low oxygen concentration for the effective functioning of the bacterial nitrogenase necessary for nitrogen fixation 115. Although the role of 2/2 Hbs in plants is completely unknown, the NsHbs, induced in plant cells upon exposure to low oxygen concentrations, are thought to play a role in a metabolic pathway which also involves nitric oxide and which provides an alternative type of respiration to the mitochondrial electron transport under hypoxic conditions 116117.

The GCSs represent one of the four known families of heme-based sensors 118, and can be subdivided into two groups, the HemATs and the gene regulators. The former are aerotactic heme sensors, with a C-terminal domain that is related to chemotaxis methyl-accepting proteins. Although Bacillus subtilis HemAT elicits an aerophilic response, that from the archaean Halobacterium salinarum provides an aerophobic response 4445. The gene regulators have C-terminal domains, some of which may regulate second messengers and others that have unknown functions [47,48.59,118]. Although little is known about the specific function of the Pgbs, they have a Cys at position E19, similar to globins of the annelids living in sulfide-rich environments, including deep sea hydrothermal vents and marine sediments, where this residue has been implicated in sulfide binding 119120.

Recent studies of Mycobacterium tuberculosis 2/2 Hbs, HbN (2/2 Hb1) and HbO (2/2Hb2), indicate that the former functions in NO detoxification, while HbO, which differs substantially from HbN in structure 121, is expressed in association with cell membranes and significantly enhances respiration, suggesting an interaction with the electron transport chain 122123124125126. It remains to be seen whether these findings apply to other bacterial 2/2 Hbs. Although nothing is known about the function of the 2/2Hb3s, it is safe to assume that it must differ from the other two classes, since 32 bacteria have two and Mycobacterium avium ssp. tuberculosis and Methylococcus capsulatus have all three classes of 2/2 Hbs.

Finally, it is appropriate to mention here the case of the extracellular Hb from the nematode Ascaris suum, an octamer of two-domain globin chains, long known for its high oxygen affinity 127, whose function appears to be that of an NO reductase 128, similar to that of fungal and bacterial FHbs. Interestingly FUGUE searches using FHb and SDgb sequences as queries, invariably include the Ascaris Hb domain 1 structure (1ash) among the highest scoring globin structures. This case provides an additional bit of evidence in support of our proposal that the FHbs and related SDgbs from bacteria, algae and unicellular eukaryotes, including fungi, are part of one of the three globin lineages, and the one from which originated all metazoan globins and all plant SHbs and NsHbs 49.

Putative globins and globin domains were identified in the genomes of 49 eukaryotes, 26 archaea and 245 bacteria, listed in Supplementary Data Table 1, using two approaches. In one, we examined the gene assignments based on a library of hidden Markov models 132, listed on the SUPERFAMILY site http://supfam.mrc-lmb.cam.ac.uk, discarded sequences shorter than 100aa and checked the alignments for the presence of His at Mb-fold position F8. In the other, we performed blastp and tblastn (version 9.2.2) searches with pairwise alignment 133, of completed and unfinished genomes in the GenBank, using the NCBI Entrez retrieval system http://www.ncbi.nlm.nih.gov/BLAST/, including the genomes of the rhodophyte (red alga) Cyanidioschyzon merolae http://merolae.biol.s.u-tokyo.ac.jp, chlorophyte (green algae) Chlamydomonas reinhardtii http://www.biology.duke.edu/chlamy/, and the diatoms Phaeodactylum tricornutum http://avesthagen.szbowler.com and Thalassiosira pseudonana http://genome.jgi-psf.org/thaps1/thaps1.home.html. Blastp searches use the Expect value (E) to assess the matches between the query sequence and each of the sequences in a database. Thus, E = 0.1 signifies that the probability of finding by chance, another match with the query sequence having the same score, is 1 in 10. We define recognition to be a hit with E < 0.005, the default threshold, and with the pairwise alignment fulfilling the following two criteria: proper alignment of the F8 His residues and of helices BC through G. It should be noted that blastp searches often misalign the E helices when the E7 residues are different, e.g. His and Q; however, the rest of the alignment is unaffected.

In cases where the identification of a putative globin was uncertain, searches employing CD-Search v.2.02 134http://www.ncbi.nlm.nih.gov, PFAM 135http://www.sanger.uc.uk and FUGUE 136http://www-cryst.bioc.cam.ac.uk were used to determine whether the borderline sequence should be accepted as a globin.

The putative globin sequences were aligned manually, using the procedure employed earlier in the alignment of over 700 globins 137, based on the myoglobin fold 138139, the pattern of predominantly hydrophobic residues at 37 conserved, solvent-inaccessible positions with mean solvent-accessible areas of <15Ų 140, including 33 intra-helical residues defining helices A through H, A8, A11, A12, A15, B6, B9, B10, B13, B14, C4, E4, E7, E8, E11, E12, E15, E18, E19, F1, F4, G5, G8, G11, G12, G13, G15, G16, H7, H8, H11, H12, H15, and H19, the three inter-helical residues at CD1, CD4 and FG4, and the invariant His at F8. Only amino acids which occur at the 33 intrahelical positions in the foregoing alignment were allowed in the alignment of the putative globin sequences. Although earlier alignments by Kapp et al. 137 and by Moens et al. 96 had indicated that there were two invariant residues in globins, F8His and CD1Phe, the 2/2Hb family can accommodate other hydrophobic residues, such as Tyr/Met/Leu/Ile/Val at the CD1 position as well as Ala/Ser/Thr/Leu at the distal E7 position, in addition to His and Gln 16202122. Hence, in our alignments, we required a His at the proximal F8 position, a residue at the distal E7 position in the order of preference His>Gln>Leu~Thr>Ala~Val~Ser~Tyr, a hydrophobic residue at position CD1 in the order of preference Phe>Tyr>Leu>Met>Ile>Val. At position C4, usually a Pro, we accepted Ala or Ser or Thr but not a charged residue. At the interhelical position CD4, we sought a hydrophobic residue, when available, and at position FG4 we placed a hydrophobic residue in the order of preference Ile>Leu~Val>Met>Phe~Tyr. Furthermore, we avoided deletions in any of the helical regions and placed no limit on the number of residues within the interhelical regions.

Bayesian phylogenetic trees were obtained employing MrBayes Version 3.1.1 141; four chains were run simultaneously for 2000000 generations and trees were sampled every 100 generations generating a total of 20000 trees. PAUP version 4.0b10 142 was used for viewing and editing. The JTT transition matrix 143 was used as the stochastic model of amino acid substitution.